MIT Department of Electrical Engineering & Computer Science

EECS Spring 1999 Catalogue Supplement

6.965 Three-Dimensional Computer Vision

WF 11-12:30, 12-142
Note: Class meets 3/1/99 - 4/16/99
Prof. Olivier Faugeras, NE43-713, x8826
Prerequisite: Basic notions of linear algebra, calculus, probabilities, random processes, image and signal processing
3-0-3

This class covers intermediate and advanced topics in three-dimensional computer vision. It is intended to provide the theoretical tools that are necessary to tackle important applications of computer vision such as the interactive and automatic modeling of three-dimensional objects and scenes, the navigation of robots, and the synthesis and coding of images. Twelve lectures will be given over six weeks. They will be partially based on the book entitled "Three-Dimensional Computer Vision: A Geometric Viewpoint," MIT Press, 1993.

I) Introduction, motivation from applications, projective geometry

II) More projective geometry and some Grassman-Cayley algebra

III) More Grassman-Cayley algebra

IV) Modeling cameras, perspective, paraperspective, affine models, nonlinear distortions

V) Traditional camera calibration: when does it fail?

VI) Systems of two cameras: epipolar geometry, fundamental matrixes, the role of planes, stereo, obstacle detection, image mosaics

VII) Representation and computation of uncertainty

VIII) Practical estimation of the fundamental matrix, outlier rejection

IX) Relation with the essential matrix, motion estimation

X) A bit more Grassman-Cayley algebra, systems of three and more cameras, trilinear relations, practical estimation techniques, new image synthesis from images

XI) Self-calibration of a camera

XII) Complete self-calibration of systems of many cameras, notions of photogrammetry, multicamera stereo, application to scene modeling